Trend or Pattern Trigger Device

Description

BACKGROUND OF THE INVENTION

1. Field of the Invention

A novel computing approach for non-systematic hypothesis generation forms the basis of a trend or pattern analyzer embodied in a device or system that typically creates an automatic triggering event, such as an alarm, telephone call, electronic transaction, etc.

2. Description of Related Art

Hypothesis generators of various kinds are known in the art—indeed, hypothesis generation is the very basis of the scientific method. Prior art hypothesis generation has inevitably been rational or at least systematic, so that a plausible potential connection between an envisionable trend and a possible outcome of that trend is first postulated and the postulation is tested. As a single illustrative example of prior art techniques, a financial professional may realize that, based on initial principles, a particular type of stock should move in a particular way. He or she then traditionally tested the idea based on empirical records (or perhaps by investing a small amount of money) to see whether the theory panned out. In other words, there has typically been a traditional path from a plausible or reasonable idea to a reasonable, tested idea which finally leads, when verifiable, to a useful and validated idea for (in this example) stock market investing. This same prior art path has traditionally played out in an unlimited number of fields in which trends are important, thus by definition including without limitation any dynamic system such as weather prediction, celestial dynamics, voter registration and election outcomes, economic forecasting including stock price changes and corporate profitability prognostics, and possibly most importantly medical diagnostics and prognostics based on patient data sets and their dynamic trends.

SUMMARY OF THE INVENTION

The present invention implements, in a trend or pattern trigger device or system, hypothesis generation while removing the heretofore unappreciated “you only ever find what you are looking for” fallacy of prior art hypothesis generation paths. The invention accomplishes this result by automating hypothesis generation with templates or charts to allow—and typically to require—consideration of trends or patterns that might or might not otherwise rationally or plausibly occur to an investigator to consider. The removal of the “you only ever find what you are looking for” fallacy is specifically accomplished by the use of trend or pattern examination templates or charts that register the presence, absence or value of nominal data and/or contain quantifiable thresholds for data or change in data without a priori specific postulation as to which data, thresholds, trends or patterns will prove important. In this way, computational analysis of data sets analyzed parameter thresholds yields trend or pattern analysis results independent of anyone's being able to, or having had to, postulate a potentially significant trend—because the computational analysis not only identifies the trend or pattern but likewise identifies which parameters or indicia are actually important. The invention is not just a computational method, however—the invention is a trend or pattern trigger device or system that takes an analytical approach applied to any given data set which then, upon identification of one or more trends, also reacts to real-world occurrence of events matching those trends by generating a further event, such as an alarm, telephone call or other concrete transaction. The implementation of this novel approach may occur in a wide variety of fields, described more particularly in the ensuing description.

DETAILED DESCRIPTION OF THE INVENTION

The present invention is in part a hypothesis generator in which templates or charts provide the thresholds for various indicia or parameters for analytical software and computational analysis to identify trends in data sets representing dynamic systems. Apart from the above general description of the inventive approach, the following initial examples is illustrative. From a financial perspective, one can use notation like Parsons codes or randomly generated charts to probe whether “stocks whose immediate past history looks like THIS (<insert chart>) will perform better/worse than average or better/worse than stocks that look like THAT (<insert chart>).” It is also possible to use financial data to perform fundamental analysis: “stocks whose value of INDICATOR is between THRESHOLD and THRESHOLD will perform better than average,” and any term in ALL CAPITALS can be filled in from a list of standard and not-so-standard financial measurements, or even randomly generated expressions or derived quantities, such as “market capitalization divided by number of employees”. The standard and not-so-standard financial measurements may or may not eliminate the seemingly ridiculous and may or may not include empirical or derived terms such as stock price divided by number of employees; sales volume divided by number of doors in factory; change in sales price divided by number of active web pages currently being maintained by the company; and even more far reaching parameters such as sales volume divided by, say, average employee age. The point of the invention is to make sure not to presuppose what the important data and thresholds are, but to investigate a wide variety of data trends or patterns in dynamic or static systems to identify, via templates and a broad if not shot-gun investigation, those trends or patterns that are statistically significant as to one or more outcomes.

At this writing, the inventors have identified approximately 3600 dynamic financial data trends that are all indicators of “buy” or “sell” for the stocks represented in the data sets. Some of these patterns (by no means the majority) are fifteen day patterns that occurred only 13 times in the particular data set. The patterns were tracked by “Parsons code” tracking, but any objective tracking system may be used. Parsons code is a tracking approach originally developed for music which allows for identification of pitch with indicators such as “up a little;” “up a lot;” “down a little;” “down a lot;” “neutral change” and so forth. More detail regarding Parsons code appears in the next paragraph. These particular fifteen day patterns of stock prices—whether up or down a little or a lot or not at all for fifteen consecutive days in the pattern identified—were strong “buy” or “sell” trend signals. However, each and every one of the approximately 3600 trends identified in the software and data set employed by the inventors—out of thousands more possible combinations—was a trend indicating a “buy” or “sell,” and thus in the context of a system or device constituted a “buy” or “sell” trigger. In other words, embodied in the system or device of the present invention, the identified trend triggers an event—an alarm, an automatic sale via electronic communication, or some other real world transformational event.

The Parsons code was technically invented for music, but can be applied to any sequence that rises and falls (like stock prices, blood glucose levels, asteroid acceleration, crop growth cycles, human IQ hour by hour throughout a single day, and limitless other examples) at discrete intervals (like musical note value lengths, or hours or days or weeks, without limitation). Each datum is either the same as the one before it (r), is higher (u), or is lower (d). For example, the song “Twinkle, Twinkle Little Star” in the key of A major has the following initial pitches: A, A, E, E, F^♯, F^♯, E. The starting note is ignored for coding purposes (which the inventors notate with an asterisk) followed by a repeat of the A, followed by an “up” transition to the E, and so forth. The final code for the above listed pitches is therefore “* R U R U R D.” A particularly useful aspect of Parsons codes is that they are scale invariant, or in other words if one wrote the song in D major instead of A major, the assigned codes would be the same. Parsons codes can also be assigned to any time values, thus making the codes potentially time invariant also. Parsons codes are also easily computable and can be easily generated at random when desired. For example, the inventors extended the Parsons code for a second set of experiments to include a richer set of transitions. Instead of just “up and “down”, the inventors allow for “up a lot” (larger than an arbitrary threshold set at 0.23%), “down a lot,” and so forth. So the final set of permitted changes is for stock price trend tracking is: *—ignore; U—up; u—up “a little;” ̂—up “a lot;” D—down; d—down “a little;” and v—down “a lot” (‘r’ is not particularly useful for stock transactions because stock prices generally don't repeat exactly). The results from the first and second experiments discussed in this paragraph appear in the attachment hereto entitled, “Exemplary Resulting Data.” The data are exemplary only and identify the trend and the analysis of the significance of the trend (including standard deviation) together with the indicator of action necessary consonant with the identified trend. More particularly, the columns in the data represent, in order, Parsons code, average % change in price, # times occurring, (average change +/standard deviation), and t-test score. The action indicator in the data governs—in the trend trigger device—the ultimate action of alarm, telephone transaction or other real world event. Please bear in mind that all the data represented on the attachment represent either “buy” or “sell” triggers—the trends that did not have enough significance to prompt an event were eliminated completely (or, more precisely, were not derived in the first place) during the computational analysis that forms a part of the present invention.

The invention is by no means limited to financial trends for stock price forecasting, however, and the invention as alluded to above is not really limited to dynamic systems at all. The template-queried data sets can analyze static systems also, such as text analysis—for authorship attribution or even for biological taxonomic or other categorization. For example, by ignoring a priori hypotheses regarding the differences between how men and women speak or write, and using the present invention, the inventors discovered that, empirically, women authors use “scent adjectives” (“fragrant,” “stinky,” “pungent” and so forth) with greater frequency than men authors do. A posteriori this finding is not surprising, because there is existing evidence that overall women have better olefactory sensitivity and acuity than men, but up until the present invention authorship attribution software did not consider or include the parameter of extent of incidence of scent adjectives as an author sex indicator. Similarly, traditional biological wisdom held that there were two species of elephant—the Asian elephant and the African elephant. Without even plumbing the genetic code of elephants, however, but just by examining various historic textual descriptions of elephants and objectively probing the patterns in the words used to describe them, it is possible to identify three trends of descriptions matching the three actual species of elephant—Asian, African Saharan and African Forest. Armed with trends identified in the textual descriptions, a researcher may then further probe the actual species' genetics for further pattern information consistent with this inventive approach. The point about authorship attribution, biological taxonomic research and other examples in this category is that they do not need to involve change over time. In the case of a static system, therefore, what would be a trend in a dynamic system is a pattern in a static system, with various indications' or parameters' being considered for pattern without the consideration of a time lapse element. Throughout this text, any reference to “trend” may be understood to refer to a pattern in a static system. Trend analysis of dynamic systems virtually always if not always includes a time lapse analysis, by contrast.

As used herein, a template or chart is a collection of one or more data field definitions in which a datum from a database is defined as to registration (presence or absence) of nominal data and/or as to threshold quantity or rate of change of quantifiable data. The template or chart is therefore useful to probe and analyze data in the database as to the presence or absence of data, various threshold quantities or rates of change (or derived quantities or rates of change) of data, and optimally there are more rather than fewer registration or threshold values in the template or chart—to reduce presumption error by having defined too few data or threshold values defined. The choice of fewer and then more threshold values is apparent in the stock price example described above and in the two sets of results shown in the accompanying attachment.

As a single non-limiting example of medical application of the present trend trigger device, the template for heart attack prediction in women over age 50, say, could include any or all of the parameters or thresholds of average blood glucose level, blood glucose level 30 minutes postprandial, average resting pulse, average blood pressure, average blood pressure on rising, length of typical work day in hours, average hours slept, genetic racial makeup, age at prima gravidae and so on. Indeed, the template parameters in this or any other example will often be dictated by the available data in the dataset—patient data that does not include length of a typical work day in hours cannot be probed for trends in which work day length is significant. However, and critically, the invention suggests that all available data be analyzed for trends rather than deciding, before analysis, which data should be considered. In the instant example, then, a patient monitor in a hospital that is cued by the software and analysis of the present invention will monitor the patient and, when the patient's indicator data patterns match a trend identified as critical by the software and device, an alarm will sound to indicate that the patient is at imminent risk of heart attack. Presumably the “alarm” is not styled to be upsetting enough to precipitate heart attack based on adrenaline surge or reaction to the alarm—instead, the “alarm” is the professional informing event or notification that motivates appropriate medical intervention, such as without limitation administration of an appropriate medicine, environmental change and/or treatment designed to intervene and to remove the patient's vital and other signs from matching the alerted trend.

The actual software or algorithm(s) used to probe and track trends or patterns are not critical in the practice of the invention. Any codes, such as the above-described Parsons code, that provide trend or pattern tracking may be used in the software portion of the present invention, and Parsons code is particularly suitable for use for actual trends that include a time element. Patterns, as opposed by trends, can be searched and matched using other software known in the art. The key to the invention is in trending or patterning a wide variety of indicia or parameters, ideally at multiple thresholds, with minimized or no human decision in advance which indicia, parameters or thresholds or even data types are important. Generally speaking, the trend or pattern trigger system or device considers and analyzes at least three indicia or parameters and more preferably at least four indicia or parameters (empiric events or derived parameters, such as stock sales price divided by number of employees). More preferably, for trend triggers the indicia or parameters are tracked and analyzed over at least four and most preferably at least five consecutive repeating time units, such as hours or days. This is not to say that an identified trend will always occur with three or more parameters or over four or more repeating time units, but the present invention will preferably consider such parameters and time units in identifying the triggering trend—which trend may be more simple than three parameters or four time units or may be much more complex.

It is also possible to implement the above analysis and trend or pattern triggering with the partial participation of a human intellect at any point throughout the process. The point of the analytical trend identification and triggering is to remove at least some of the a priori human prejudice from the trend identification, if not all of it.

Finally, it should be borne in mind that the present invention has application to virtually any database. Typically in the prior art, databases are queried with the desired output in mind. In the present invention the software and computational analysis look for trends or patterns in all available data as tied to an outcome (stock price increase or decrease, imminent onset of heart attack, etc.) By extension, then, it will always be possible to design a database as to which the computational analysis of the invention is deployed. In keeping with the idea of the invention, database design of this type is best conducted in an inductive way—including all available data types or fields rather than trying to rationalize which data fields will prove important. This implementation requires a change of mindset from traditional database design. Heretofore, data was included in a database in fields already determined to be important and relevant. In the present invention, however, virtually always any database should be compiled to include all available data or derived data—for example not just the batting histories and ages of the players on the Little League team, say, but also each player's hair color, shoe size, grade point average, runs batted in divided by years on the League, and whatever else in the way of data and thresholds is available. In other words, in order to implement the present invention all potentially available data is analyzed for trends or patterns, not just the data a priori believed to be important. At a macro level, the software and computational analysis of the present invention becomes a hypothesis generator in a way that substitutes for human hypothesis generation. At a practical, real world level the invention institutes pattern or trend identification without limiting human prejudice and deploys such pattern or trend identification to match real world patterns or trends to prompt action in the event of a pattern or trend match. Such a trend or pattern trigger device or system can thus intelligently govern buy or sell orders in an investment portfolio, predict and precipitate the averting of asteroid strikes, or prevent heart attacks or strokes—to name just a few.